How to Fix Subtitle Sync Problems: Delay, Drift and Frame Rate Mismatch
Subtitles early, late or drifting? Diagnose the cause and fix it with a simple offset, a two-point correction or a frame rate conversion.
People use the words captions and subtitles interchangeably, and most of the time nobody is confused. But when you are making files for real audiences, particularly for accessibility, the difference matters. A file made for viewers who can hear the audio and need a translation is not the same as one made for viewers who cannot hear it at all.
This guide explains the vocabulary, shows what each type contains, summarises what accessibility guidance expects, and gives practical tips for making caption files that really serve the people who rely on them.
Subtitles assume the viewer can hear the audio but does not understand the language. They translate (or transcribe) the dialogue, and usually nothing else. If a door slams or music swells, the subtitle does not mention it, because a hearing viewer already knows.
Captions assume the viewer cannot hear the audio, either because they are deaf or hard of hearing, or because the sound is off. Captions include the dialogue, who is speaking when it is not obvious, and meaningful non-speech sounds such as music, laughter, and sound effects.
SDH stands for Subtitles for the Deaf and Hard of Hearing. It is the term often used on discs and streaming services for subtitles that contain caption-style information. In practice an SDH file is a caption file delivered in a subtitle format and, frequently, in the same language as the audio.
The terms also vary by region. In North America, caption is the usual word for same-language text for deaf and hard-of-hearing viewers, and subtitle usually means translation. In much of Europe and elsewhere, subtitle is used for both.
Another pair of terms describes whether the viewer can switch the text on and off.
Closed captions are a separate layer that the viewer can toggle. They match what we have called soft subtitles in our comparison of soft vs hardcoded subtitles.
Open captions are burned into the picture and cannot be turned off.
Both can contain the same words. The difference is only in how they are delivered.
Dialogue is only part of the job. Captions for people who cannot hear the soundtrack need to carry the information that the sound carries.
When the speaker is off screen, or when several people could be talking, add the name or a label. Common styles are a name in capitals followed by a colon, or the name in square brackets: [MARIA]. Use one style consistently. When the speaker is clearly visible, a label is unnecessary.
Put meaningful sounds in square brackets, in present tense: [door slams], [phone ringing], [crowd cheering]. Describe only sounds that matter to the story or the mood. A caption for every footstep makes the text tiring to read.
Indicate music and, when it matters, its mood or lyrics. Many styles use a musical note symbol for sung lyrics and a bracketed description for background music, such as [tense music].
Words alone can mislead if the delivery changes their meaning. A short cue such as [sarcastically] is appropriate when the tone is essential to understand the scene.
Use italics for voices that come from outside the scene, such as a phone call or narration, if your style guide does. Two speakers talking at once can be shown with two lines and a dash for each speaker.
Here is the same moment as a plain subtitle and as an SDH caption.
Plain subtitle:
1
00:00:12,000 --> 00:00:14,500
Did you hear that?
SDH caption:
1
00:00:11,200 --> 00:00:12,000
[glass shatters]
2
00:00:12,000 --> 00:00:14,500
MARIA: Did you hear that?
The caption adds a sound that the viewer cannot hear and a speaker label for someone who may be off screen.
Accessibility rules vary by country and by context, so check the requirements that apply to you. A widely used reference is the Web Content Accessibility Guidelines (WCAG), which ask for captions for prerecorded video that includes audio at the basic conformance level. Many laws and procurement policies refer to WCAG, so captioning is often an obligation for organisations, not an optional extra.
Some broadcast regulators go further and define caption quality. For example, the standards used for television captions in the United States focus on four ideas: accuracy, synchronicity, completeness, and placement. These are good principles for any caption project:
Rules and thresholds change, so use the current official documents for any formal requirement and treat this article as an overview.
Accessibility features often help far beyond their original audience. People watch videos with the sound off on public transport and in offices. People in noisy rooms, people learning a language, and people watching in a second language all use captions. Text also makes video content easier to search and quote. Producing good captions is rarely wasted effort.
Start with a transcript. Automatic speech recognition can produce a first draft quickly. It must be reviewed by a human, because names, jargon, accents, and punctuation are frequently wrong, and it knows nothing about sounds that are not speech.
Use SRT or WebVTT for delivery. Both are widely accepted by video platforms and players. Our comparison of subtitle formats explains which to choose.
Follow reading-speed rules. Captions are often verbatim, which makes them fast. Use the guidance in subtitle formatting best practices: short lines, sensible durations, and a reading speed that suits your audience. For some content, light editing for readability is better than a strict word-for-word transcript.
Review with the sound off. Watch the video without audio and check whether you can follow the story from the captions alone. If you cannot, a sound description or speaker label is probably missing.
Compare drafts. When an automatic draft and a human edit both exist, a side-by-side tool such as the SRT Merger lets you accept the edits line by line, keeping accurate lines from the draft and the improved ones from the human version.
Treating subtitles as captions. A translation-only subtitle file leaves out the information that deaf and hard-of-hearing viewers need.
Over-describing sounds. Too many bracketed notes bury the dialogue. Keep to what matters.
Inconsistent labels. Switching between several speaker-label styles looks careless and slows reading.
Ignoring placement. Captions that cover on-screen text or faces fail the placement principle. Move them when necessary, in a format that supports positioning.
No review. An unchecked automatic caption file can contain embarrassing and misleading errors.
Subtitles translate speech for people who can hear. Captions convey speech and meaningful sound for people who cannot, and SDH is captions delivered as subtitles. Closed means switchable, open means burned in. Good caption files identify speakers, describe important sounds, keep a readable pace, and are accurate, synchronised, complete, and well placed. Check the accessibility rules that apply to your audience, review every automatic draft by hand, and keep an editable SRT or WebVTT file as your source.
Merge your own subtitle files
Compare two .srt files side by side and save the best lines as one file. It runs in your browser, so nothing is uploaded.